English negative questions: A view on acquisition at the syntax-discourse-prosody interface

We present two elicited imitation studies investigating the morphosyntax and prosody of three types of negative questions in British English (BrEng). Children imitated negative polar questions (NegQs), negative tag questions (TagQs) and “why-don’t-you” questions (WDQs); positive polar questions (PosQs) acted as controls. In BrEng, TagQs are more frequent in input than, and accurate in child output before, NegQs (Woods/Roeper 2021, Heim 2023). If frequency matters, we expect elicited TagQs to be more accurate than NegQs both in syntax and prosody. WDQs are ‘frozen’ forms common in child input/output (Kania 2016), so we anticipate accurate imitation; we also predict accurate imitation of PosQs.

In study 1, 31 children (3;5-8;1, mean 5;5) produced 492 questions imitating a robot puppet with monotonous prosody. Ages >5;3 (n=18) made only 9 morphosyntactic errors in 286 utterances; the error rate for ages <5;3 (n=14) was 94/222 (42%). In ages <5;3, TagQs gave rise to most errors (n=35/55), followed by WDQs (24/57), NegQs (23/54), and PosQs (12/56); the main error was change of polarity. Non-adult-like questions+tags also resulted where a TagQ’s declarative anchor was converted to an interrogative (n=8). Prosody typically remained robotic.

In pilot study 2, 7 children (3;11-6;6) produced 115 questions imitating a cat avatar with a human voice. PosQ, NegQ and WDQ morphosyntax was accurate at >80%, with TagQs at 58% accuracy (n=12/29); the main error type was dropping negation in tags (n=6). For prosody, imitation of placement of the prenuclear accent on the auxiliary was accurate at ceiling in NegQs, with low error rates in TagQs/WDQs. More errors occurred in PosQs, where accenting of the auxiliary is less common in the input. Nuclear contour imitation accuracy dropped in NegQs (70%) and WDQs (71%); common errors include replacing (fall-)rises with falls (n=12) or shallower rises (n=10).

 Input frequency cannot fully explain our results. The pragmatically richer context of study 1 may have mitigated against WDQs, and we will suggest that TagQ errors are due to underlying syntactic complexity rather than recall issues. Prosodic errors in NegQ nuclear contours reflect variation in the adult input even at age 6. The distinct WDQ contour errors remain puzzling.

 References: • Heim (2023). Acquisitional trajectories of English tags, Paper at LAGB 2023. • Kania (2016). The Acquisition and Use of Yes-no Questions in English: A Corpus-Study from a usage-based perspective, Narr Francke Attempto Verlag. • Woods & Roeper (2021). The acquisition path of ‘high’ negation in English. BUCLD 45, 761–775.